Possession and the Standings: Explaining the Playoffs

This week we looked at whether looking at a team’s possession numbers in the regular season (in conjunction with the regular season points) helped explain playoff performance. Even if you’re already sold on the importance of puck possession, it’s not obvious that this should be the case. Puck possession has already been established at being very important in determining regular season standings, so you might think that, after 82 games, the standings already reflect pretty much everything you need to know about a team’s possession game. In other words, you might expect that the inclusion of possession metrics into a regression wouldn’t find much of an effect because the effect of possession (and more) is already captured in the regular season standings.

As it turns out, this isn’t the case.

To begin with, we looked at the probability of winning a playoff series as a function of only the regular season point differential. For this, we used all playoff series going back to the 1997-98 season. As we have done in the past, we ran a probit regression, for which an explanation can be found here. Our findings were that each additional point a team had over its opponent increased its probability of winning by about 1.1 percentage points.

Unfortunately, we don’t have possession data going back to 1997-98, so we then had to restrict our sample to start with the 2007-08 season. Counting the first 2 rounds of this year, that gave us 102 series, which is not a huge number, but not small either. We considered two different measures for possession, Fenwick For % (FF%) and Corsi For % (CF%), both in 5v5 situations with the score close. The data came from Hockey Analysis. The regression results can be found here.

The results from the regression using FF% are reported in The Star: each point difference in the regular season leads to an increase in the probability of winning the series of 0.6 percentage points; and each percentage point difference in FF% in the regular season leads to an increased probability of winning the series of 2.5 percentage points. The regression using CF% is slightly different: the effect of point differential is about 2/3 of a percentage point, and the effect of Corsi differential is 2 percentage points.

The drop in the effect of point differential is stemming from the correlation between possession and regular season standings. The fact that the effect of possession metrics are significant and somewhat sizable tells us that they offer some explanatory value over and above how they affect the standings.

There are a couple of possible explanations for this. First, it is possible that the true quality of a team is still not fully captured by the standings after 82 games. That is, `puck luck’ could still be big enough a factor after 82 games compared to the importance of possession. Another possible explanation is that a style of playing the game that focuses on possession leads to greater success in the playoffs than it does in the regular season. There is nothing in the regressions we have run that can distinguish between these two possible explanations, although it should be possible for us to come to a conclusion on this issue with further analysis. Of course, there may be other explanations as well.

No Comments Yet

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>